Light Field Synthesis by Training Deep Network in the Refocused Image Domain

TL;DR

This paper introduces a novel deep learning approach for light field synthesis that optimizes the quality of refocused images directly, improving the visual quality of synthesized light fields for applications like image refocusing.

Contribution

The paper proposes a new refocused image error (RIE) loss function that enhances light field synthesis by focusing on refocused image quality, a novel approach in the domain.

Findings

Improved refocused image quality over previous methods.

Better performance on real and synthetic datasets across multiple metrics.

Effective in balancing spatial and angular resolution trade-offs.

Abstract

Light field imaging, which captures spatio-angular information of incident light on image sensor, enables many interesting applications like image refocusing and augmented reality. However, due to the limited sensor resolution, a trade-off exists between the spatial and angular resolution. To increase the angular resolution, view synthesis techniques have been adopted to generate new views from existing views. However, traditional learning-based view synthesis mainly considers the image quality of each view of the light field and neglects the quality of the refocused images. In this paper, we propose a new loss function called refocused image error (RIE) to address the issue. The main idea is that the image quality of the synthesized light field should be optimized in the refocused image domain because it is where the light field is perceived. We analyze the behavior of RIL in the spectral domain and test the performance of our approach against previous approaches on both real and software-rendered light field datasets using objective assessment metrics such as MSE, MAE, PSNR, SSIM, and GMSD. Experimental results show that the light field generated by our method results in better refocused images than previous methods.